Hazard-map colors can't choose a battery-storage site── Measuring siting viability and business continuity with Expected Annual Loss (EAL)

"Passed over because it falls inside an inundation hazard area" — this is the single most common way a grid-scale storage site dies in screening. But a hazard map is a life-safety (evacuation) map under the Flood Control Act, and what it draws is the worst-case image at an annual exceedance probability of 1/1000 or less. You cannot measure a 20-year business with a map that has erased the frequency information. The government has already built probability maps for business decisions (multi-stage inundation maps); combined with the Flood Control Economic Survey Manual, a location hazard can be translated into Expected Annual Loss (EAL) — a monetary figure. From a 2MW/8MWh model calculation, to the cost-effectiveness of raising the site, to the ripple into insurance ("how you cover it and what it costs," alike), and on to a 14-item workflow that reads one candidate site in a single day — this piece builds the yardstick that comes after the binary decision, using nothing but public primary information.

The scope of this piece — translating "colored or not" into "a monetary figure and insurance terms."

01A hazard map is a "life-safety map"

The inundation depths drawn on a municipality's flood hazard map assume, as a rule, the maximum assumed scale (L2) — rainfall with an annual exceedance probability of roughly 1/1000 or less (MLIT, Manual for Preparing Flood Inundation Hazard Area Maps, 4th edition ✅). This does not mean "a once-in-1,000-years cycle"; it means "each year, the probability of exceeding that scale is 0.1% or less." The purpose is to secure evacuation under the Flood Control Act. Because human life must not be lost even in a once-in-a-millennium flood, the map displays the single worst image — and as institutional design, that is correct.

The problem arises when that "single worst image" is repurposed as a binary filter for an investment decision. The color for 3 m of inundation looks the same whether it is a 1/1000 story or a 1/30 story, yet the expected loss differs by two orders of magnitude. Judging a 20-year operating business with a map that has erased the frequency information is a mismatch of probability dimensions.

And this repurposing strikes straight at the land inventory. Follow the ideal siting conditions for grid-scale storage — close to the connection point, flat, cheap land — and candidates cluster in alluvial lowlands, and most of Japan's alluvial lowlands take on some color at L2. Hold the line of "colored = out of contention," and candidates drift toward hillsides and terraces, where you now trade for a different set of costs: sediment-disaster risk, earthwork, and connection distance. A binary decision does not eliminate risk. It merely reassigns it to a different risk.

02The government also provides "maps for business decisions"

The probability data needed for the re-reading all exist as public primary information.

Probability (flood) ✅: Since December 2022, MLIT has been rolling out Flood Risk Maps (multi-stage inundation maps). These show the inundation depth of the same land in multiple stages by annual exceedance probability — 1/10 (high frequency), 1/30 (mid-high), 1/50 (mid), 1/100 (mid-low), planning scale (1/150 or 1/200), and maximum assumed scale — and are a different object from the single L2 image. Decisively, the preparation guideline itself (Water and Disaster Management Bureau, January 2023) states verbatim that the uses of this map include "corporate site selection," "corporate BCP (business continuity plan) preparation," and "calculation of flood-insurance rates" (✅ confirmed in the body text). "There is only a life-safety map" is already a thing of the past.

Probability (earthquake): NIED's J-SHIS provides exceedance probabilities such as "seismic intensity 6-lower or greater within 30 years" on a nationwide 250 m mesh. Climate-change adjustment: under a 20-year operating premise, it is reasonable to fold in, as a sensitivity, the official coefficients from MLIT's proposal to revise flood-control planning (October 2019 / revised April 2021 ✅) — under a 2 °C scenario, rainfall ×≈1.1, river discharge ×≈1.2, flood frequency ×≈2.

And this very re-reading into "probability × damage amount" is an evaluation framework the government has been operating on real rivers. In the Kinokawa river system, for example, MLIT's Kinki Regional Development Bureau (Wakayama River and National Highway Office) publishes multi-stage inundation maps and flood risk maps (listed in MLIT's "Flood Risk Map Index" ✅); and for the same Kinokawa's directly-managed river-improvement project, the Bureau's project evaluation (cost-benefit analysis) computes and publishes damage-prevention benefit — an expected damage amount with the same skeleton as this piece's EAL — from the annual average of households and area whose inundation is mitigated, based on the Flood Control Economic Survey Manual (Project Evaluation Monitoring Committee materials). The government actually uses this yardstick to decide flood-control investment. What we have independently researched and systematized is the connecting joint that re-reads this governmental evaluation framework into the investment and insurance decisions for a private asset — a storage facility.

03From "does it fall inside?" to "how much will you lose?" ── building the EAL

Expected Annual Loss (EAL) = Σ (annual-exceedance-probability band × damage rate × asset value) + Σ (annual-exceedance-probability band × effective outage days) × daily lost revenue − expected insurance recovery

The first term is asset impairment; the second represents business continuity — how many days it stops, and how you bridge revenue in the meantime. The guideline lists "BCP preparation" among the map's uses precisely because of this second term. For damage rate and outage days, there are official governmental benchmarks.

※1 Flood Control Economic Survey Manual (draft), April 2005 edition, Table 4.4, depreciable assets (✅ confirmed matching the source table). Damage rates were updated in the April 2020 edition; for actual projects we recommend referring directly to the latest April 2024 edition itself. ※2 April 2020 edition, Table 4.9 (business-suspension days + business-slowdown days × 1/2; ✅ recomputed to match source figures). ※1 and ※2 are based on editions from different revision years.

But these are nationwide averages for a general business establishment, and BESS needs two adjustments — both of which push the numbers in the harsher direction.

Adjustment ①: read the damage function as a step function (inundation = 100% total loss)

For homes and businesses the damage rate rises gradually with depth, but BESS is different. Battery modules, PCS, and receiving/transforming gear are electrical and electronic equipment set at ground level; because of stranded energy, a submerged lithium-ion battery is hard to discharge or repair safely, and with salt water or contaminated water there is a path from terminal corrosion to internal short and thermal runaway. Commentary from overseas BESS-specialist insurers (e.g., Altelium) treats an inundated battery unit as, typically, ending up a total loss, and the manufacturer's warranty IP rating (IP65, etc.) covers dust and splashing — flooding is outside the warranty scope and can be a warranty-voiding cause. The manual itself, taking the automobile as the representative asset with an electrical system, sets 100% total loss at 70 cm of inundation (seat level) or more (April 2020 edition ✅). If water reaches the equipment foundation plane, 100%; if not, essentially 0% — this step shape is the technically honest damage function.

And this step shape is exactly where design becomes the pressure point. Deciding the raise height is nothing other than buying, with money, "up to which frequency on the exceedance-probability curve you zero out damage." There are private-sector implementations. The UK's Harmony Energy operated Pillswood BESS (98MW/196MWh, live November 2022, among Europe's largest at commissioning ✅) was built on an earthwork platform precisely because it sits on a flood-risk site — an operator buying out the left side of the probability curve with money and turning colored land into an operating asset. Note that water doesn't only stop revenue; it also starts fires — the ignition of PG&E's Elkhorn Battery (September 2022) was identified as rainwater entering the container due to faulty installation of a deflagration vent shield panel (PG&E disclosure ✅).

Adjustment ②: outage duration is governed by equipment procurement lead time

The manual's outage days (≈98 days at 3 m+) are actuals for a general business, and in a BESS total loss this becomes the floor. Receiving/transforming equipment is made to order (MLIT/METI's Guideline on Flood Countermeasures for Building Electrical Equipment likewise notes that recovery requires a substantial period). Extra-high-voltage transformer lead times run 128 weeks on average (≈2.5 years) on an actual-order basis, and ≈143–144 weeks for step-up GSUs (Wood Mackenzie 2025 Q2 survey, U.S. T&D market ⚠️) — and the structure is the same at home: made to order, long lead time. PCS runs 6–12 months, and battery containers on the order of 4–6 months from a practical standpoint (estimates).

On top of the outage sit market penalties. In the capacity market, once the annual cumulative of requirement-missed slots exceeds 8,640 slots (equivalent to 180 days), it becomes subject to an economic penalty. And outages whose capacity-outage plan was not submitted by the prescribed deadline (e.g., 17:00 the prior Tuesday), and missed slots that fell during a tightness event with a wide-area reserve ratio below 8%, can be counted at 5× as missed slots (OCCTO Capacity Market Operations Manual, target delivery FY2026 ✅). "Days stopped" cut into revenue by a route separate from the damage amount, too.

04A model calculation ── putting a number on "a riverside site"

This is a model calculation we built independently. The premise: HV 2MW/8MWh, ¥400M depreciable asset (CAPEX ≈ ¥480M = ¥60,000/kWh). That unit cost is the analytical reference value from ANRE's Study Group on Expanding Stationary Storage Systems; the FY2024 actuals run around ¥54,000 system + ¥14,000 installation per kWh (✅). Daily lost revenue is ¥130k/day (capacity market + JEPX arbitrage; the annual gross of "≈¥46M" is rounded on an effective-utilization basis — at 365-day full operation it is ≈¥47.45M). For inundation probability, we placed as an illustration a riverside site read from a multi-stage map as "inundation begins at 1/10, 0.5 m at 1/30, 1 m at 1/50, 2 m at 1/100, and 3 m at maximum assumed scale (1/1000)."

Probability bands are illustrative (for a real site, read the actual figures from that river's multi-stage inundation maps). Calculation by ScienceX (checked by hand).

The result: physical-damage EAL ≈ ¥15.6M/yr + outage EAL ≈ ¥0.33M/yr ≈ ¥16M/yr total. Cumulatively over 20 years (undiscounted), ≈ ¥320M, equal to ≈67% of CAPEX (even at a 5% discount, present value ≈ ¥200M ≈ 40% of CAPEX). The intuition that "a riverside lowland is tough as a business" is borne out even when quantified — up to here, the conclusion is the same as a binary decision.

The difference shows in the next move. The same yardstick can also measure the price of countermeasures. Raise the equipment foundation plane by 1 m, and the probability bands at 1 m depth or less — the thickest part of the exceedance-probability curve — are zeroed out, and EAL nearly halves to ≈ ¥7.2M/yr (a reduction of ≈ ¥8.8M/yr; because the effective inundation depth against the equipment also drops by 1 m in the upper bands, the actual reduction is larger still). Not "avoid it because it's colored," but "compare the raising cost against the ≈ ¥8.8M/yr reduction — and against the improvement in insurance terms discussed below" — this decision becomes possible. Try it in the calculator below by moving asset value, daily revenue, and raise height.

Three cautions, stated plainly. First, the probability bands are illustrative. Second, the table's damage rates are general-business values; apply the BESS adjustment (reaching equipment level = 100% total loss) and they push higher when equipment sits below the raise height — meaning the damage rate is also a variable you can move by design, not just by location. Third, the outage EAL is clearly a floor (with a total loss of the receiving/transforming gear, the order of magnitude of outage days changes).

05Property insurance changes ── the ripple into "how you cover" and "what it costs"

An EAL evaluation is a tool for the investment decision and, at the same time, an input to insurance design. This is the biggest practical difference from a binary decision.

Insurers read the same map. For retail fire insurance, the flood rate was subdivided from October 2024 into five bands by municipality (GIROJ, filed June 2023 ✅; the inter-regional gap is ≈1.2× on the total premium and ≈1.5× on the flood rate alone). That is a retail reference net rate and corporate properties are individually underwritten, but the direction — reflecting location risk into the rate — is the same. Overseas, BESS-specialist underwriters go further still: specialist commentary (e.g., Altelium) holds that color-coded information such as "falls in a 1-in-100-year zone" is insufficient for a location evaluation, and requires inundation depths (meter values) at multiple return periods. The side that graduated from the binary first is the side that underwrites the insurance. Conversely, an operator who can present its own probability-differentiated inundation depths and countermeasure design can negotiate terms at the same resolution as the underwriter. A bad map cannot be arbitraged with insurance — it comes back as premium and deductible — but good data becomes leverage that moves the terms.

On that basis, EAL prescribes the specifics of insurance placement (how you cover it).

• Structure of flood cover: in corporate property insurance, flood is covered with a payout ratio (selectable, e.g., 5–70%), a deductible, and a payout limit. At a site where raising has zeroed out the high-frequency bands, the small-and-frequent claims vanish, so it becomes rational to take a thicker deductible to hold down the rate and concentrate cover on the tail (low-frequency, large loss) — the idea being to align the EAL curve with the choice of deductible and payout ratio.
• Whether an earthquake endorsement is needed: under policy wording, earthquake, eruption, and tsunami are standard exclusions, and tsunami is classified on the earthquake side, not as "flood." Liquefaction and ground subsidence, if earthquake-triggered, also fall on the same excluded side. Filling this requires an earthquake-risk endorsement (extended cover), but its underwriting availability, cover-ratio cap, and rate are non-public individual quotes (❓ no public data). The more a site's hazard composition leans to the earthquake side (tsunami, liquefaction), the more the "uninsured without the endorsement" residual piles up — and this residual cannot be erased in insurance design; it can only be controlled at the site-selection stage.
• Indemnity period for BI (business interruption): a ~1-year indemnity period is the practical starting point, but a total-loss transformer lead time is on the order of 2.5 years (U.S. survey figure). Lost revenue after the indemnity period is exhausted is structurally uninsured. Whether and at what cost to extend the indemnity period should be weighed on the same field as a spares strategy (spare transformer, etc.) (BI-design specifics in COLUMN 18).
• Parametric as a complement: products that pay early, without loss adjustment, triggered by an index such as seismic intensity, actually exist domestically. Tokio Marine & Nichido's "EQuick earthquake-ready insurance" launched in March 2021 as Japan's first index-type product (✅ per the company's own release). Mitsui Sumitomo's seismic-intensity-linked endorsement (2025), Swiss Re's payout roughly two months after the 2022 Fukushima offshore earthquake, and the entry of overseas players such as Descartes are per each company's disclosures and reporting (⚠️). It is not full-loss indemnity and basis risk remains, but it can function as initial liquidity for recovery funds that compensates for the long loss adjustment of BI.
• Timing-of-payment risk: BESS fires and accidents tend to have protracted cause investigations. In the storage-facility fire in Isa City, Kagoshima (March 27, 2024), despite the facility being insured, the payout was reported still outstanding more than a year after the event (PVeye June 2025 issue ✅), and as of July 2026 we have not confirmed any follow-up report of payment. Insurance is evaluated not by "is it in place?" but by "when, how much, and on what conditions does it pay out?"

In sum, EAL is a common denominator that compares, in the same currency, the land-price reduction, countermeasure cost, premium, and self-retained residual. The question shifts from "should we insure it?" to an allocation problem: "which loss layer do we pay with engineering, which with insurance, and which with the land price?"

06The three-tier sort ── veto / priceable / design input

A building-collapse inundation zone is a veto because it is not a function of inundation depth. This zone is the area where, when a levee breaches, the fluid force of the flood flow and bank erosion are expected to topple or sweep away structures (two types — flood flow and bank erosion; publication advanced under the Flood Control Act in the wake of the 2015 Kanto–Tohoku heavy rain ✅). The damage mode is not "submersion" but "swept away, foundation and all," which raising does not erase; the damage rate is effectively 100%, and insurance underwriting breaks down on terms because of the location itself. Quantification is a tool to "save colored land," and at the same time a tool to show, in the investment committee's language, the land that cannot be saved.

Liquefaction is, in principle, priceable. A hazard map's "high liquefaction potential" verdict (PL value, etc.) is an evaluation of the original ground and does not fold in the effect of ground improvement. In other words, liquefaction can be priced as a ground-improvement cost. But the residual risk falls, as noted, on the earthquake-exclusion side — and this is the decisive difference from flood. The market tends to treat liquefaction lightly not because the risk is low, but because reading it takes geotechnical literacy. The developer/reseller holding period is short, and it is the long-term holder that owns the site until the seismic risk materializes — so for the buyer, in particular, it is a point worth scrutinizing.

Near a fault is, except directly overhead, a design input. The purpose of building-setback regulation for active faults is life-safety in residential buildings. In representative confirmable examples, California uses 50 feet (≈15 m) from the fault (Alquist–Priolo Act ✅), while New Zealand, Taiwan, and domestic district-plan examples are cited at the 10–25 m class (a rough figure per the Japanese Geotechnical Society Chubu Branch's summary, etc. ⚠️). Damage from a surface fault rupture concentrates in the strip directly overhead (in the 1971 San Fernando earthquake, building damage was ≈80% directly overhead versus under 30% in the offset zone ✅); for an unmanned container structure, surface displacement is avoided by pad placement, and shaking is handled as a foundation and anchor design condition — which is engineering-accurate. Here again, this piece's principle repeats: "do not extend life-safety maps and regulations to the business decision of an unmanned facility."

07Precedents, and the named evaluators ── next to seismic PML, the flood-version seat sits empty

The vocabulary for "accepting institutional-investor capital in a country with hazards, nonetheless" has been built over a quarter-century by the market next door: seismic PML in real-estate securitization (probable maximum loss at a 475-year return period). A working convention has taken hold as a practical benchmark — treat roughly 15–20% PML as a threshold, and above it, consider or make a condition of earthquake insurance (⚠️ the threshold range is industry practice). No one says "because it's an earthquake-prone country, real-estate investment is impossible."

And the practical evaluation infrastructure is already coming together on the storage side, too. Tokio Marine Group's Tokio Marine dR publishes facility-risk reports (natural disaster, fire) for grid-scale storage (Tokio dR-EYE ✅) and, holding that a quantitative natural-disaster evaluation is indispensable at the time of project-finance structuring for renewable generation facilities, offers PML evaluation commercially (per its own service page ✅; the flood evaluation combines inundation hazard-area maps + Fathom's return-period flood hazard + inundation simulation). Sompo Risk Management likewise offers probabilistic seismic/tsunami PML evaluation that computes loss amounts by return period, and flood-risk evaluation. In third-party technical DD, TÜV Rheinland Japan has offered technical due diligence for grid-scale storage (including location evaluation and site survey) since November 2023, and as of October 2024 it discloses that it has completed several domestic planning/development-stage projects, with reports frequently requested not only by development operators but by banks and financial institutions considering PF (✅ per its own press release). Among listed infrastructure funds, per-property seismic-PML disclosure of held generation assets and portfolio PML management are already established practice.

In short, the evaluating side (insurer-affiliated risk consultancies, third-party certification bodies) and the vessel for disclosure (fund practice) already exist. What cannot be confirmed in public information is only an operator-side disclosure that an individual storage project was selected on multi-stage probability × EAL (❓ DD reports are non-public by convention). Conversely, there is not yet a seller who hands this statement to investors on its own — that is where the first-mover room lies. Turning the annual exceedance probabilities of the multi-stage map, the manual-conformant damage rate (with BESS adjustment), the outage days based on procurement lead time, and the insurance cover scope and exclusions into a single EAL statement changes the one-line "because it's riverside" into negotiable questions: "is it dry to equipment level at 1/100?" and "what percent of annual revenue is the residual EAL?"

08The limits of this framework

Expected-loss evaluation is not a cure-all. To run it honestly, we list up front the holes that cannot be filled.

• Structural holes in insurance: with payout ratio, deductible, and payout limit, even flood does not return in full. Earthquake, eruption, and tsunami are standard exclusions, and extended-cover terms are property-dependent (❓ no public standard).
• Payment delay: with a protracted cause investigation, the receipt of proceeds and the demand for recovery funds can be out of phase (the Isa City example).
• Time does not come back: insurance returns the physical asset, but not the re-procurement period of made-to-order equipment, nor lost revenue beyond the indemnity period.
• In-period rate variation: the BESS insurance market has a shallow loss history, and there is a risk of terms worsening at renewal.
• Premise obsolescence: hazard maps and earthquake probabilities are updated. The Nankai Trough 30-year occurrence probability changed in the September 26, 2025 revision to a dual notation of "roughly 60–90% or more" and "20–50%" (base date January 1, 2025 ✅; because the base date is updated each January, always check the latest value at the time of evaluation). The risk that "outside the zone" at acquisition becomes "inside the zone" in the future is an uncertainty the EAL model itself carries.
• Blanks in public information ❓: the rate level of BESS property + BI insurance, the underwriting reality of earthquake extended cover, and the practice of the BI indemnity period and lost-profit calculation — none are publicly available. They can only be filled by a direct inquiry to insurers, which, conversely, means whoever holds the inquiry results stands at an information advantage.

09Practical workflow ── Monday morning, when one candidate-site address arrives

That's the logic. The practice folds into four steps. The data used in ① is, registry aside, all free public information.

① In the first 30 minutes eliminate veto-class; in one day, produce the full picture of the candidate

The entry point is the Overlay Hazard Map. Enter the address and first look at whether it falls in a building-collapse inundation zone (flood flow / bank erosion) or a Special Landslide Hazard Zone. If it does, stop there — spend no effort on ② onward. Surviving candidates make one pass through the following 14 items. This list is to avoid both "look only at hazard and die on regulation" and "look only at regulation and die on hazard."

The example URLs are the prefectural/city systems for the Kinokawa basin (Wakayama) and Okayama — for city planning, forests, embankments, and fire ordinances there is no single nationwide portal, and the systems are split by municipality. In other prefectures too you can reach the same kinds of system via "prefecture name + integrated GIS" and "municipality name + city-planning map." Two reading cautions. First, the regulatory items (8–10) bite with the same weight as a hazard veto — a blue zone, a protected forest, a control area without review criteria are "time-out-type" stoppers that cannot be priced like an EAL, and if you don't kill them at stage ①, the quantitative evaluation of ② onward is wasted. Second, this screening is done not to exclude but to line up the variables to pass to ②. As needed, it's good to check, on the same day, buried-cultural-property zones (the municipality's ruins map; Okayama Prefecture Integrated GIS also carries a buried-cultural-property layer) and the stormwater-runoff-control obligation (in many municipalities, a site over 1,000 m² is the rule of thumb for a detention pond).

② One Excel sheet is enough for the EAL

The §04 table becomes the template as-is. The key is to hold "the planned height of the equipment-foundation plane (GL + how many m)" as a variable — simply set the bands where the by-probability inundation depth reaches the foundation plane to 100% damage, and those that don't reach it to 0%, and you have the BESS-version EAL as a step function. If, here, there is no prospect of the residual EAL coming within a single-digit % of annual revenue and no room for countermeasures, it is a sound decision to withdraw rather than proceed to ③.

③ Back-calculate the raise height from "the frequency you protect"

Rather than thinking about "how many meters to raise" first, decide first "up to which annual exceedance probability you zero out damage" (e.g., the 1/100 inundation depth + a freeboard at the foundation top surface; the idea is the same as the high-installation and flood-defense line in the Guideline on Flood Countermeasures for Electrical Equipment). The means are three: whole-site embankment fill (public unit-cost rule of thumb ≈ ¥6,500–7,400/m³; retaining wall ¥20,000–80,000/m²), raising the equipment stand/foundation (partial and cheaper; prioritize raising equipment that is "totaled if wet," such as receiving/transforming and PCS), and pile foundations (which double as anchorage to the bearing stratum and a liquefaction countermeasure). Where liquefaction is also present, ground improvement — surface improvement ≈ ¥3,000–7,000/m², column-type ¥10,000–30,000/m², and the whole countermeasure ¥10,000–100,000/m² (falling as the treated area grows) are the public unit-cost rules of thumb. But these are centered on public unit costs for housing and general civil works, and some are direct-construction costs as of 2011 — with recent construction-price inflation they run high, so always confirm the BESS-scale actuals with an EPC / ground-survey quote (⚠️). The yardstick for the decision is always "countermeasure cost (annualized) vs. EAL reduction" — in the §04 example, raising by 1 m corresponded to a reduction of ≈ ¥8.8M/yr. Being able to make this comparison at all is graduation from the binary decision.

④ Bring "meter values," not "colors," to the insurer

The inquiry package is three items: inundation depths by probability, the foundation-top height, and the countermeasure design drawings. The answers to obtain are three: (a) the flood payout ratio, deductible, payout limit, and rate; (b) the underwriting availability and terms of the earthquake-risk endorsement (the top-priority inquiry item at a site where tsunami/liquefaction weigh heavily); and (c) the BI indemnity-period cap and waiting period. The uninsurable / high-deductible portions are not treated as vanished — they remain, as monetary figures, in the EAL statement as the "self-retained residual." What you ultimately bring to the investment committee is one set: the EAL statement + countermeasure cost + insurance terms + the three-condition check that follows. On a deal with a lender attached, obtaining third-party technical DD (including location evaluation) ahead of time lets you pre-empt the credit side's questions.

◆Three conditions to bring to the investment committee

Waiting for "land with no flood risk whatsoever" is, in Japan's flatlands, synonymous with forgoing the opportunity. Instead, we propose sorting deals by the following three conditions.

• ① The EAL based on the by-probability inundation depths of the multi-stage inundation map (with damage rates corrected to a step function, 100% for total loss) comes within the business plan's tolerance — as a rule of thumb, the residual EAL within a single-digit % of annual revenue.
• ② Flood cover (with payout-ratio and deductible design included) and, where the hazard composition requires it, an earthquake-risk endorsement stand at an economically rational rate.
• ③ It does not fall in a veto-class such as a building-collapse inundation zone or a Special Landslide Hazard Zone.

These three conditions rewrite the definition of "investable land." Even land that is colored at L2 — if it is dry to equipment-installation level up to 1/100 on the multi-stage map, and the insurance structure closes without excessive reliance on an earthquake endorsement — can be a superior site with a slim residual EAL. Conversely, a lowland that submerges under high-frequency inland flooding is precarious even if lightly colored. Where the line should be drawn is not the presence of color, but the level of the EAL.

A hazard map is not a map that tells you whether to invest. It is a map to protect human life, and for the business, it is input data for expected loss. Japan's largest inventory of suitable land — the alluvial plain — opens, in order, to those who hold the vocabulary for that re-reading.